Can digital labour market platforms get Africa’s youth working? Evidence from Mozambique

The intervention is for a randomised sample of technical and vocational school graduates (who are being tracked as part of an on-going school-to-work transition survey) to be sent an SMS inviting them to register either with one of two job sites (two treatment arms). The first job site caters for workers to advertise their services in manual/practical skills (Biscate), and the second is a more traditional job site, where employers advertise job vacancies (Emprego).

2020-03-01

2020-03-31

The primary outcome is whether or not the individual is in work

The primary null hypothesis is that usage of digital labour market platforms does not alter labour market outcomes. The corollary one-sided alternative hypothesis is that usage of said platforms does improve labour market outcomes.

Measures of platform usage, other measures of economic activity, and, at other measures around the final outcome.

Obviously, a necessary condition for the experiment to be able to identify the contribution of platforms to labour market outcomes is that the SMS invitations (treatments) increase registration on and subsequent use of the platforms. This therefore represents a subsidiary hypothesis to be tested. Schematically, the potential relationship is that treatment will affect platform usage, which will affect economic activity. While economic activity might itself be considered a final outcome (e.g., whether or not unemployed) it is helpful to make a distinction between activity and other outcomes. For instance, an individual already employed on a full-time basis might use the platforms to find more lucrative employment or negotiate better working conditions. In this case, while one would not find any impact of the platforms on the rate of economic activity (or being employed), earnings or other aspects of employment would change.
Measures of platform usage:
• Separate dummy variables for whether or not the individual is registered on Emprego
or Biscate; data obtained directly from UX and cross validated with phone interview data.
• Dummy variable for whether the individual has used digital platforms to search for work over the past week (aggregated over the rounds)
• Intensity of activity on Emprego or Biscate, as measured by the number of times the user uses the job search site (Emprego, page views on Google Analytics) and number of contacts from clients / confirmations of work agreements made (Biscate); data from UX.
2. Other Measures of economic activity
• Dummy variable for whether the individual has undertaken work (performed some kind of economic activity) in the past seven days, or states they have a regular job (e.g., on holiday).
• Time in months looking for work (out of employment) since graduation.
• Number of hours worked in main activity in past week.
3. Measures of final outcomes
• Total income after tax in past month (inverse hyperbolic sine transformed)
• Income after tax from primary work in past month (inverse hyperbolic sine transformed)
• Income after tax from secondary work in past month (inverse hyperbolic sine transformed)
• Number of jobs obtained via Biscate; job obtained from organization on Emprego in response to application
• Index of job quality, constructed as the count of the following characteristics – having a permanent or fixed position, having a written contract, being registered in the social security system (INSS), working the desired number of hours (neither over or under employed), whether the individual is actively seeking another job (conditional on being employed), working in the same area of one’s studies; and working in a job in which technical qualifications are necessary to perform required tasks. (The latter measures of horizontal and vertical mismatch may also be analysed separately).
• Overall subjective match of current work situation to original expectations as at baseline survey (dummy for better than expected).

The experiment investigates the role of digital labour market platforms in matching workers and employers. The primary research question, therefore, is: can digital labour market platforms help young people in Mozambique find work?
We investigate the contribution of two distinct platforms, one which allows workers with practical/manual skills to advertise their services, and the other is a more traditional job website, where employers post vacancies.
To identify whether these platforms contribute to finding work and improved job outcomes we will adopt an ‘encouragement design’, where we will send SMS invitations to participants selected randomly from the baseline sample to register either with Biscate or with Emprego (two treatment arms).
This RCT will be embedded within the ongoing tracer study and, as such, will leverage an on-going data collection (quarterly phone survey) which is collecting a wide range of information from each participant, including their current work status, type of work undertaken, conditions of employment (e.g., with/without a contract), the number of hours worked, and income earned from employment (in primary and secondary activities).

The two treatment arms relate to two distinct job-seeking platforms, both operated by the same technology company (UX). Each website targets a very different segment labour market, and also facilitates matching by a different process.
1. Biscate: (biscate.co.mz) where individuals with practical/manual skills can advertise their availability and expand their base of customers for what is primarily occasional ("odd-job") work. The platform runs primarily via mobile USSD (Unstructured Supplementary Service Data) on the Vodacom network (only), but also has a dedicated App and website (www.biscate.co.mz). Currently, the platform has around 50,000 registered workers.
2. Emprego (emprego.co.mz): is a more traditional jobs website where employers post vacancies and can receive applications from registered users. The platform is only accessible via the internet, either using an App or the website (www.emprego.co.mz). Currently, the site is accessed by more than 18,000 individuals daily and over 1,400 organizations in Mozambique use the service.

The SMS invitations sent to the the two treatment arms (for Biscate or Emprego) will be personalized (based on their name and area of study) and follow-up reminders will be sent after 1 week to individuals who have not joined the platform to which they were invited.
Considering that the Emprego invitation is generally most relevant (but not exclusively) to graduates of services-oriented courses, while the Biscate invitation is generally more relevant to graduates of manual (technical) courses, we apply a conditional or partial randomization rule, whereby we randomly assign a moderately larger proportion of graduates from manually-oriented courses to Biscate and a moderately larger proportion of graduates from services-oriented courses to Emprego.
After treatment, platform usage will be monitored, and economic activity will be tracked, both by collecting data from the platforms (e.g. number of log-ons, number of vacancies applied to, number of jobs agreed) as well as quarterly phone surveys. The analytical focus will be on the endline results (after one year), incorporating information from the full tracking period.

A conditional or partial randomization rule

Individual

No

1,433 individuals (recent graduates)

1,433 individuals (recent graduates)

604 Individuals in the control group
396 Individuals invited to join the Emprego job site (treatment arm 1)
433 Individuals invited to join the Biscate job site (treatment arm 2)

Unit = share of individuals working, Percentage = 0.3, Standard Deviation = 0.75, Power = 0.7955 Based on the selected sample size of 1,433 and using an alpha of 10% and a beta of 80%, we find an MDE of 0.0755 for the Emprego treatment (treatment arm 1) and 0.0735 for the Biscate treatment (treatment arm 2).

Design and Pre-analysis Plan document